Back to Business & Economics

CoreWeave funds its AI buildout with customer prepayments and shorter, pricier contracts

CoreWeave reported $3.66 billion in third-quarter revenue on Tuesday, up 168% from the year before, along with a $709 million net loss and $846 million in quarterly interest expense. The neocloud, which rents out the GPUs behind AI, used the earnings to describe a change in how it pays for growth: it is increasingly asking customers to pay up front, borrowing against specific contracts, and taking on shorter-term deals that carry higher prices.

A $132.5 billion backlog, funded a different way

The numbers frame the problem. CoreWeave's revenue more than tripled year over year, but it still posted a net loss, and it spent $846 million just on interest in the quarter. That is the cost of financing a backlog of $132.5 billion in committed contracts.

CEO Michael Intrator put the financing decision plainly on the earnings call: "We will not use the unsecured debt market to fund our business at current yields, nor do we need to." The company's plan is to lean on loans that are backed by specific customer contracts, drawing funds as it needs them to build out the infrastructure those contracts require.

CoreWeave said it expects roughly $9 billion in customer prepayments and deposits over the next three months. More than $3 billion of that is tied to signing new contracts for capacity the company has already built, which means no additional hardware spend to deliver it.

The prepayment trade-off

Upfront money eases the near-term borrowing load. But it also locks CoreWeave into providing services at a price that could turn out to be unattractive if compute demand keeps climbing. A customer who pays today for capacity priced at this quarter's rates gets to keep that rate.

That is the tension at the center of the new model. CoreWeave is trying to harvest cash from demand that is, right now, strong. If that demand is durable, the prepayments are a clean source of funding. If it is a spike, the company has already sold capacity at prices the market may soon leave behind.

Shorter deals, higher prices

The second lever is contract length. CoreWeave said it is allocating more capacity to short-term opportunities where the pricing justifies it, selling at more than $40 million per megawatt annually, while pricing for new infrastructure on four- and five-year contracts is moving to $20 million to $25 million per megawatt.

Short-term deals let the company charge more and reset prices more often. The numbers back it up: CoreWeave raised prices across its offerings by about 25% in July and by another 10% later in the quarter, and it said prices are still climbing enough to "more than offset" rising costs. In August it closed a $2.6 billion loan, its first built specifically for shorter-term customer contracts. Known as DDTL 5.5, it lets CoreWeave borrow in installments as it needs the money.

The catch is that lenders weight longer contracts more heavily, even at lower prices, because they value predictable cash flow. Short-term paper is juicier for the company and thinner for the banks.

What the bond market thinks

Wall Street took the earnings well. Bond investors were less sure. Last month, yields on some of CoreWeave's bonds climbed above 15%, a sign that markets are wary of the debt being piled up to fund data center buildouts before those investments pay back.

The company's rivals point the same way, in the sense that they are making the same move. Nebius and IREN have both leaned on large prepayments from Microsoft to fund their own capacity, which suggests the whole neocloud layer is shifting the bill onto Big Tech customers.

For a team that buys AI compute, the read is concrete. The cheapest access may be getting tied to longer commitments or to paying ahead of delivery, while short-term capacity is being reserved for the buyers willing to pay a premium. CoreWeave is betting that the demand that made its revenue grow 168% is the kind that holds.

Source: techcrunch.com

More news

Back to Business & Economics

Have something in mind?

Start a conversation